Vera Rubin NVL72 tokens/s per megawatt vs GB200 NVL72 — CoreWeave measured ~10× on DeepSeek-R1 at a matched interactivity target on first live silicon; NVIDIA frames it 'up to 10×' (ratio only, no absolute tokens/sec or $/token disclosed)
~10× (up to 10×)observed
| Value kind | observed — Reported measurements, counts, and specifications keep the precision and scope stated by their source; an exact specification is not treated as a range. |
|---|---|
| Caveat | Power-normalized capacity density at matched latency on one workload (DeepSeek-R1); not a proven fleet-wide energy or cost cut. Absolute tokens/sec, rack kW, and $/token are not disclosed. |
| As of | 2026-07 |
| Source | CoreWeave |
| Review | checking…review by 2026-09-23 · fast cadence |
| Recorded changes | last 2026-08-24 · 2 revisions tracked |
| Claim id | vera-rubin-nvl72-tokens-per-megawatt-vs-gb200 |
Where the guide uses it
Not quoted in a chapter yet; it is kept in the curated register.
← Full numbers register — every date-stamped figure in the guide, with revision history.